Upload
atreyib
View
7.524
Download
3
Tags:
Embed Size (px)
Citation preview
Role of Programming in Computational
Biology
Atreyi Banerjee
Programming languages
• Self-contained language– Platform-independent– Used to write O/S– C (imperative, procedural)– C++, Java (object-oriented)– Lisp, Haskell, Prolog (functional)
• Scripting language– Closely tied to O/S– Perl, Python, Ruby
• Domain-specific language– R (statistics)– MatLab (numerics)– SQL (databases)• An O/S typically manages…
– Devices (see above)
– Files & directories
– Users & permissions
– Processes & signals
Role of Programming
Reduces time Reduces money on R&D Reduces human effort Streamlines workflow Standard Algorithm unifies data result and
assures reproducability Reduces human error
Applications of Programming Data Mining Genome Annotation Microarray Analysis Website Development Tool Development Statistical Analyses Phylogeny Genome Wide Association Studies (GWAS) Next Generation Sequencing studies
Bioinformatics “pipelines” often involve chaining together multiple tools
Perl is the most-used bioinformatics language
Most popular bioinformatics programming languages
Bioinformatics career survey, 2008
Michael Barton
PERL Practical Extraction & Report Language Interpreted, not compiled
Fast edit-run-revise cycle
• Procedural & imperative Sequence of instructions (“control flow”) Variables, subroutines
• Syntax close to C (the de facto standard minimal language)
Weakly typed (unlike C) Redundant, not minimal (“there’s more than one way to do it”) “Syntactic sugar”
High-level data structures & algorithms– Hashes, arrays
Operating System support (files, processes, signals) String manipulation
Pros and Cons of Perl
• Reasons for Perl’s popularity in bioinformatics (Lincoln Stein)– Perl is remarkably good for slicing, dicing, twisting, wringing, smoothing,
summarizing and otherwise mangling text
– Perl is forgiving
– Perl is component-oriented
– Perl is easy to write and fast to develop in
– Perl is a good prototyping language
– Perl is a good language for Web CGI scripting
• Problems with Perl– Hard to read (“there’s more than one way to do it”, cryptic syntax…)
– Too forgiving (no strong typing, allows sloppy code…)
General principles of programming
Make incremental changes Test everything you do
the edit-run-revise cycle Write so that others can read it
(when possible, write with others) Think before you write Use a good text editor Good debugging style
Regular expressions
Perl provides a pattern-matching engine Patterns are called regular expressions They are extremely powerful
probably Perl's strongest feature, compared to other languages
Often called "regexps" for short
Programming in PERL
Data Types Scalars ($) Arrays (@) Hashes (%)
Conditional Operators AND (&&), OR (||), NOT (!)
Arithmetic Operators (+, -,*, /)
CONDITIONS If else Elsif ladder
LOOPS For While Foreach
Default Variables $_ default variable @_ default array
Finding all sequence lengthsOpen file
Read line
End of file?
Line starts with “>” ?
Remove “\n” newline character at end of line
Sequence name Sequence data
Add length of line to running totalRecord the name
Reset running total of current sequence length
First sequence?Print last sequence length
Stop
noyes
yes
yes
no
no
Start
Print last sequence length
DNA Microarrays
Normalizing microarray data
• Often microarray data are normalized as a precursor to further analysis (e.g. clustering)
• This can eliminate systematic bias; e.g. if every level for a particular gene is elevated, this might
signal a problem with the probe for that gene if every level for a particular experiment is elevated, there
might have been a problem with that experiment, or with the subsequent image analysis
• Normalization is crude (it can eliminate real signal as well as noise), but common
Rescaling an array
For each element of the array:add a, then multiply by b
@array = (1, 3, 5, 7, 9);print "Array before rescaling: @array\n";rescale_array (\@array, -1, 2);print "Array after rescaling: @array\n";
sub rescale_array { my ($arrayRef, $a, $b) = @_; foreach my $x (@$arrayRef) { $x = ($x + $a) * $b; }}
Array before rescaling: 1 3 5 7 9Array after rescaling: 0 4 8 12 16
Array is passed by reference
Microarray expression data A simple format with tab-separated fields First line contains experiment names Subsequent lines contain:
gene name expression levels for each experiment
* EmbryoStage1 EmbryoStage2 EmbryoStage3 ...Cyp12a5 104.556 102.441 55.643 ...MRG15 4590.15 6691.11 9472.22 ...Cop 33.12 56.3 66.21 ...bor 5512.36 3315.12 1044.13 ...Bx42 1045.1 632.7 200.11 ...... ... ... ...
Messages: readFrom(file), writeTo(file), normalizeEachRow, normalizeEachColumn…
Reading a file of expression datasub read_expr { my ($filename) = @_; open EXPR, "<$filename"; my $firstLine = <EXPR>; chomp $firstLine; my @experiment = split /\t/, $firstLine; shift @experiment; my %expr; while (my $line = <EXPR>) { chomp $line; my ($gene, @data) = split /\t/, $line; if (@data+0 != @experiment+0) { warn "Line has wrong number of fields\n"; } $expr{$gene} = \@data; } close EXPR; return (\@experiment, \%expr);}
Note use ofscalar contextto comparearray sizes
Reference to array of experiment namesReference to hash of arrays(hash key is gene name, arrayelements are expression data)
Normalizing by gene A program to normalize expression data from a
set of microarray experiments
Normalizes by gene
($experiment, $expr) = read_expr ("expr.txt");while (($geneName, $lineRef) = each %$expr) { normalize_array ($lineRef);}
sub normalize_array { my ($data) = @_; my ($mean, $sd) = mean_sd (@$data); @$data= map (($_ - $mean) / $sd, @$data);}
NB $datais a referenceto an array
Could also use the following:rescale_array($data,-$mean,1/$sd);
Normalizing by column
Remaps gene arrays to column arrays
($experiment, $expr) = read_expr ("expr.txt");my @genes = sort keys %$expr;for ($i = 0; $i < @$experiment; ++$i) { my @col; foreach $j (0..@genes-1) { $col[$j] = $expr->{$genes[$j]}->[$i]; } normalize_array(\@col); foreach $j (0..@genes-1) { $expr->{$genes[$j]}->[$i] = $col[$j]; }}
Puts columndata in @col
Puts @colback into %expr
Normalizes (note use of reference)
Genome annotations
GFF annotation format• Nine-column tab-delimited format for simple annotations:
• Many of these now obsolete, but name/start/end/strand (and sometimes type) are useful
• Methods: read, write, compareTo(GFF_file), getSeq(FASTA_file)
SEQ1 EMBL atg 103 105 . + 0 group1SEQ1 EMBL exon 103 172 . + 0 group1SEQ1 EMBL splice5 172 173 . + . group1SEQ1 netgene splice5 172 173 0.94 + . group1SEQ1 genie sp5-20 163 182 2.3 + . group1SEQ1 genie sp5-10 168 177 2.1 + . group1SEQ2 grail ATG 17 19 2.1 - 0 group2
Sequencename
Program
Featuretype Start
residue(starts at 1)
Endresidue
(starts at 1)Score
Strand(+ or -)
Codingframe
("." if notapplicable)
Group
ArtemisA tool for genome annotation
Reading a GFF file
• This subroutine reads a GFF file• Each line is made into an array via the split command• The subroutine returns an array of such arrays
sub read_GFF { my ($filename) = @_; open GFF, "<$filename"; my @gff; while (my $line = <GFF>) { chomp $line; my @data = split /\t/, $line, 9; push @gff, \@data; } close GFF; return @gff;}
Splits the line into at most ninefields, separated by tabs ("\t")
Appends a reference to @datato the @gff array
Writing a GFF file
• We should be able to write as well as read all datatypes• Each array is made into a line via the join command• Arguments: filename & reference to array of arrays
sub write_GFF { my ($filename, $gffRef) = @_; open GFF, ">$filename" or die $!; foreach my $gff (@$gffRef) { print GFF join ("\t", @$gff), "\n"; } close GFF or die $!;}
open evaluates FALSE ifthe file failed to open, and$! contains the error message
close evaluates FALSE ifthere was an error with the file
GFF intersect detection
• Let (name1,start1,end1) and (name2,start2,end2) be the co-ordinates of two segments
• If they don't overlap, there are three possibilities:• name1 and name2 are different;•
name1 = name2 but start1 > end2;
• name1 = name2 but start2 > end1;
• Checking every possible pair takes time N2 to run, where N is the number of GFF lines (how can this be improved?)
Self-intersection of a GFF file
sub self_intersect_GFF { my @gff = @_; my @intersect; foreach $igff (@gff) { foreach $jgff (@gff) { if ($igff ne $jgff) { if ($$igff[0] eq $$jgff[0]) { if (!($$igff[3] > $$jgff[4] || $$jgff[3] > $$igff[4])) { push @intersect, $igff; last; } } } } } return @intersect;}
Note: this code is slow.Vast improvements inspeed can be gained ifwe sort the @gff arraybefore checking forintersection.
Fields 0, 3 and 4 of the GFF line are the sequence name, start and end co-ordinates of the feature
Converting GFF to sequence
• Puts together several previously-described subroutines• Namely: read_FASTA read_GFF revcomp print_seq
($gffFile, $seqFile) = @ARGV;@gff = read_GFF ($gffFile);%seq = read_FASTA ($seqFile);foreach $gffLine (@gff) { $seqName = $gffLine->[0]; $seqStart = $gffLine->[3]; $seqEnd = $gffLine->[4]; $seqStrand = $gffLine->[6]; $seqLen = $seqEnd + 1 - $seqStart; $subseq = substr ($seq{$seqName}, $seqStart-1, $seqLen); if ($seqStrand eq "-") { $subseq = revcomp ($subseq); } print_seq ("$seqName/$seqStart-$seqEnd/$seqStrand", $subseq);}
Phylogenetics Analysis of relationships between organisms
through phylogenetic programs like PHYLIP can be automated or run on command line
Packages
Perl allows you to organise your subroutines in packages each with its own namespace
Perl looks for the packages in a list of directories specified by the array @INC
Many packages available at http://www.cpan.org/
use PackageName;PackageName::doSomething();
This line includes a file called"PackageName.pm" in your code
print "INC dirs: @INC\n";
INC dirs: Perl/lib Perl/site/lib .The "." means thedirectory that thescript is saved in
This invokes a subroutine called doSomething()in the package called "PackageName.pm"
Object-oriented programming
Data structures are often associated with code FASTA: read_FASTA print_seq revcomp ... GFF: read_GFF write_GFF ... Expression data: read_expr mean_sd ...
Object-oriented programming makes this association explicit.
A type of data structure, with an associated set of subroutines, is called a class
The subroutines themselves are called methods A particular instance of the class is an object
OOP concepts
• Abstraction– represent the essentials, hide the details
• Encapsulation– storing data and subroutines in a single unit– hiding private data (sometimes all data, via accessors)
• Inheritance– abstract base interfaces– multiple derived classes
• Polymorphism– different derived classes exhibit different behaviors in response
to the same requests
OOP: Analogy
OOP: Analogy
o Messages (the words in the speech balloons, and also perhaps the coffee itself)
o Overloading (Waiter's response to "A coffee", different response to "A black coffee")
o Polymorphism (Waiter and Kitchen implement "A black coffee" differently)
o Encapsulation (Customer doesn't need to know about Kitchen)
o Inheritance (not exactly used here, except implicitly: all types of coffee can be drunk or spilled, all humans can speak basic English and hold cups of coffee, etc.)
o Various OOP Design Patterns: the Waiter is an Adapter and/or a Bridge, the Kitchen is a Factory (and perhaps the Waiter is too), asking for coffee is a Factory Method, etc.
OOP: Advantages
• Often more intuitive– Data has behavior
• Modularity– Interfaces are well-defined– Implementation details are hidden
• Maintainability– Easier to debug, extend
• Framework for code libraries– Graphics & GUIs– BioPerl, BioJava…
OOP: Jargon
• Member, method– A variable/subroutine associated with a particular class
• Overriding– When a derived class implements a method differently from its parent
class
• Constructor, destructor– Methods called when an object is created/destroyed
• Accessor– A method that provides [partial] access to hidden data
• Factory– An [abstract] object that creates other objects
• Singleton– A class which is only ever instantiated once (i.e. there’s only ever one
object of this class)– C.f. static member variables, which occur once per class
Objects in Perl
• An object in Perl is usually a reference to a hash• The method subroutines for an object are found in a
class-specific package– Command bless $x, MyPackage associates variable $x with package MyPackage
• Syntax of method calls– e.g. $x->save();– this is equivalent to PackageName::save($x);– Typical constructor: PackageName->new();– @EXPORT and @EXPORT_OK arrays used to export
method names to user’s namespace
• Many useful Perl objects available at CPAN
Common Gateway Interface
• CGI (Common Gateway Interface)– Page-based web programming paradigm
• Can construct static (HTML) as well as dynamic (CGI) web pages.
• CGI.pm (also by Lincoln Stein)– Perl CGI interface– runs on a webserver– allows you to write a program that runs behind a
webpage
• CGI (static, page-based) is gradually being supplemented by AJAX
GUI
Graphical User Interface (GUI) are standalone modules created to make the work of an end user simpler.
Can be achieved through PERL Tk
BioPerl• Bioperl is a collection of Perl modules that facilitate the
development of Perl scripts for bioinformatics applications.
• A set of Open Source Bioinformatics packages– largely object-oriented
• Implements sequence and alignments manipulation, accessing of sequence databases and parsing of the results of various molecular biology programs including Blast, clustalw, TCoffee, genscan, ESTscan and HMMER.
• Bioperl enables developing scripts that can analyze large quantities of sequence data in ways that are typically difficult or impossible with web based systems.
• Parses BLAST and other programs• Basis for Ensembl
– the human genome annotation project– www.ensembl.org
Basic BioPerl modules
Bio::Perl Bio::Seq Bio::SeqIO Bio::Align Bio::AlignIO Bio::Tools::Run::StandAloneBlast
BLAST
CLUSTALW
References
http://www.bioperl.org http://www.cpan.org http://www.pasteur.fr Learning Perl Beginning Perl for Bioinformatics Mastering Perl for Bioinformatics
Thank You